Covid-19 forecaster errors wrecked Govt decision-making

By Simon Thornley, Gerhard Sundborn, Ananish Chaudhuri and Michael Jackson.

It is clear now that estimates of death from the Covid-19 pandemic were exceeded by factors of hundreds, if not thousands. This sparked public and political panic and led to our government enacting one of the most stringent lockdowns in the world.  Te Pūnaha Matatini predicted 80,000 deaths even with mitigation strategies, while the University of Otago team forecast 12,600 to 33,600 deaths.  Their best possible estimate was 5,800 deaths. The models encouraged the government to enact tight control measures. Now, we are largely over the epidemic, although some of the modelers have warned of secondary waves. New Zealand now has 22 ‘official’ Covid-19 deaths – a far cry from the forecast doom and gloom, with at least a 263 fold over estimate at this point. A recent article about Sweden followed suit, predicting a total of 60,000 deaths for that country, and decrying its decision not to lockdown.

How was it possible for these forecasts to be so erroneous? The interesting aspect, reading the modelling now, is that the number infected under each control policy scenario, including lockdown, was about the same. The Matatini group described 89% of the population being ultimately infected under even the most stringent strategy. The moment the handbrake was let off, another outbreak would occur. However, in the paper, the modellers themselves questioned the effect of lockdowns. They wrote:  “In other countries, including those that have instigating (sic.) major lockdowns such as Italy, there is as yet insufficient evidence that this has reduced [the epidemic]”. They then stated that “successful mitigation requires periods of these intensive control measures to be continued for up to 2.5 years before the population acquires a sufficient level of herd immunity.” The conclusion was that lockdowns were buying time for vaccination and learning from other countries. The modelling that justified the lockdowns was itself clearly stating that such policies were far from a panacea.

Models incorporated lockdown measures yet still predicted thousands of deaths. Critics will say that the lockdown is precisely why the models were so inaccurate. We were saved from catastrophe. Several lines of consistent statistical evidence does not, however, support this idea. US States that did not lockdown report lower Covid-19 cases and death rates on average than States that enforced heavier restrictions. Time trends in Europe show that lockdowns prolonged the recovery from the epidemic after these policies were enforced. Closer to home, it is clear that cumulative per capita cases and deaths of Covid-19 are lower for Australia than for New Zealand despite more relaxed restrictions over the Tasman.

The major factors behind these erroneous models include: (1) an overestimate of the infection fatality rate, and (2) a reciprocal underestimate of the immunity of the population.  Mathematical models of infections project the assumptions of the modellers into the future. They are mathematically elegant, but also based on many untested assumptions. Models assume a far greater degree of certainty than is true in reality.

The models used are built for infections which declare themselves, like measles. Covid-19 is different, it produces high rates of infections in people who feel well. Measles primarily affects young children who are unlikely to die from other causes. Covid-19, on the other hand, has shown to be most vicious at the other end of the age spectrum, specifically causing death most frequently in people at a mean age very similar to our life expectancy, about 82 years. This is curious, as it strongly suggests that the virus does not shorten life, since our life expectancy, or average lifespan, is similar with or without the virus on board. There is little mention of this in the Matatini document, and it is relegated to the appendix of the University of Otago report. Instead the Otago group talk of deaths of the magnitude seen in World War I. Given the age differences of deaths in World War I (mean about 27 years), compared to Covid-19, this must surely be classed as exaggeration.

Neither modelling team attempted to quantify loss of life in terms of ‘years of life lost’ (YLL), a standard epidemiological technique for comparing disease burden. Such statistics would have produced a totally different picture than headline death tallies, portrayed simplistically by the media. YLL is the sum of the differences between age at death and median life expectancy and weights death in the young higher than deaths in the old. Since Covid-19 deaths occurred at an average age in the 80s, this method of measurement would have produced a much less striking picture than the less sophisticated count that values infant and nonagenarian mortality as equivalent. Years of life lost from Covid-19 are extremely low, and pale in comparison to other risks to health, such as cardiovascular disease, diabetes and cancer.

As in the case of swine flu, antibody tests of the virus, are dialling down the infection fatality rate, to a range similar to influenza (0.03% to 0.5%). This contrasts from the genetic test evidence used by some commentators. This cuts down the dire predictions for Sweden by a large ratio. Since even people without antibodies have evidence of seeing the virus, the true infection fatality ratios are likely to be even lower than those adjusted for antibody tests alone. It is now clear that the dire prediction is very unlikely to be correct, since Sweden is now well into the downward slide of its epidemic curve for Covid-19 deaths (figure). The value of observed data over modelled predictions is demonstrated here.

Figure1 (above). Epidemic curve of Covid-19 deaths in Sweden (1/June/2020). Line represents average trend.

Related to the immunity tests, a strong, and very questionable assumption of the modelling is that we are all, as a population, susceptible to the ‘novel’ virus. Since from early on in the epidemic, it was clear that infection was more likely in the elderly, this was unlikely to be so. Recent evidence from immunologists strongly indicate cross-reactivity between “common cold” coronaviruses and SARS-CoV-2, which was present in at least 30% of people that don’t show other evidence of having seen the disease before. This theory is supported by a study that showed that 34% of a sample of healthy blood donors who did not have antibodies, instead had other evidence of immunity, with reactive T cells to the virus. Also, the finding of test-positive samples in France well before the epidemic ‘officially’ occurred, dents the ‘we are all sitting ducks’ theory.

In trying to make sense of these erroneous predictions we have to ask how this happened? We believe two basic features of the human psyche have been at work. The first of these is loss aversion: the desire to avoid losses that are right in front of us even if it means larger losses elsewhere or further down the road. The second is confirmation bias: that is the tendency to look for evidence that confirms one’s pre-supposition and discounts evidence that calls those beliefs into question. Of course, the 24-hour news-cycle, the cacophony of social media, the need for eyeballs, clicks, likes, tweets and retweets exacerbates these matters, since apocalyptic predictions are more likely to draw attention.

Several lines of evidence give us hope, to counter pessimistic modelling. One thing the inaccuracy of the models teach us is that our understanding of the behaviour of the virus is incomplete. Better understanding should translate to more accurate prediction. Epicurves by country in Europe and many parts of Asia, along with Australia and New Zealand are showing waning epidemics with insignificant secondary peaks. These patterns strongly suggest growing immunity in these countries, despite measured low antibody prevalence in some areas. The high rates of cellular and cross immunity explains this phenomenon. China, a very densely populated country, has now widely opened up after a lockdown and had few secondary waves. Japan is the same, although they had lighter restrictions. The sustained low number of cases when the curve falls strongly indicates that we can safely return to normality much more rapidly than was thought possible.


Norway officially concludes that its lockdown was not necessary

the Norwegian public health authority has published a report with a striking conclusion: the virus was never spreading as fast as had been feared and was already on the way out when lockdown was ordered.

“It looks as if the effective reproduction rate had already dropped to around 1.1 when the most comprehensive measures were implemented on 12 March…”

Are the coronavirus epidemiological models any good?

Health issues India discuss coronavirus models with Dr Simon Thornley.


Why the prejudice against tests for Covid-19 immunity?

Simon Thornley


Words: 1090

A curious phenomenon has developed in the race to beat Covid-19. Advisors to the government have recently become anti anti-bodies. Before I explain what that means, let me provide some context. While we’ve weathered the initial Covid-19 storm, we now have a more challenging set of questions ahead of us as we decide how far and fast to ease social restrictions and open our borders back up to the world.

One of the most critical is: just how widespread is this virus? If, as the Government’s advisors believe, it’s a case of ‘what you see is what you get’, then our options are limited. But if, as we are seeing around the world, the virus has spread through far more of our population than we are aware, then that changes everything. All of a sudden, we need to radically re-think whether our control measures make sense. The genetic test that we are relying on can tell us if the virus is active in the here and now. That is the focus of the daily case counts. These tests are accurate, and the best for diagnosing cases, but they don’t give us a complete picture.

In almost all infectious diseases, antibody tests play a crucial role in determining who is protected from the germ and who is not. They tell us that a virus or germ has been and gone. They are the fingerprints that the virus leaves behind, and allow us to be better prepared for the next encounter. For Covid-19, we may not otherwise know we have met and dispatched the virus, since not all of us develop symptoms. In Iceland, of the few areas of the world a survey was carried out, rather than only testing sick people, 1% of the population tested positive, but half all these positives were perfectly well. It is now clear that just because we don’t have a fever, runny nose or cough, it doesn’t mean we haven’t seen the virus. For this reason, we simply cannot rely on genetic tests from people with symptoms to tell us how far the virus has spread. To really get a handle on how many of us have seen a virus, we need to not only count active cases, but start measuring people who have seen the virus before with antibodies.

New Zealand is now at a cross-roads. We have two explanations for our results. Professor Michael Baker, one of the main experts advising the government, has expressed that antibody tests “would be a waste of time and resources” since a “vanishingly small” proportion of the population have been exposed. Through Baker’s eyes, the lockdown was astonishingly effective, quashing the virus, while leaving all except the one and a half thousand or so cases sitting ducks waiting for infection to strike. We had better live in fear and shut down the borders hard. This narrative goes with the elimination story. So much for our travel and tourist industry. Sorry Rotorua and Queenstown, we have laid you on the altar as a casualty on the path to vanquishing the virus.

Another explanation for the rise and fall of cases in New Zealand is from growing immunity, rather than from the lockdown. The cases of infection rise as the virus encounters more susceptible people. This is great for the virus until it encounters people who have seen the virus before. Their bodies have wised up, thanks to our miracle antibody factories, and the virus sees the door is shut. Some may not even need antibodies. The innate and cellular immune system, like a razor wire fence, may keep the virus out before the soldier-like antibodies need to be enlisted.

Immunity from other viruses is also likely to play a part. A recent study estimated that half of people who haven’t seen the novel virus before, have T cells that react against it which are primarily directed against ‘common cold’ coronaviruses. The virus looks elsewhere, but the door is shut with the next person, and the next, and it soon has nowhere to go. This has been the way we have defeated almost every other lung virus of equivalent severity to Covid-19 in the past.

Now critics will say there are holes in this immunity theory. If that had really happened, we should have seen chocka intensive care units like in Italy. Well, we may have, or we may not. It is clear that New Zealand is not Milan, London and New York, as we would like to believe. We are simply nowhere near as population-dense as these metropolises.

Surely we would have noticed excess deaths? Or excess people coming to hospital with influenza-like symptoms? Since the deaths from Covid-19 are about the same average age as our life expectancy, we may not have noticed. If we hadn’t tested for it, we would have probably not batted an eyelid. We would have put the death down to the growing list of diseases that were likely to have afflicted the deceased. And it is not as if Covid-19 gives a unique clinical presentation. As a former hospital doctor, I know only too well that patients who present with flu-like illness are extremely common. A recent positive test in a French patient well before the ‘official’ epidemic occurred support this theory of widespread infection.

Teasing out which of these two beliefs to follow is now critical. History may help. In recent memory, a story played out according to the widespread immunity theory. We strongly believed that H1N1 was a killer virus, rapidly spreading out of Mexico. The death rate was astonishingly high initially. The clamour to ‘stamp out’ the virus in New Zealand was long and loud. It was, at least, until needles were put in veins, and antibodies were present in 47% percent of some age groups. These tests established that many New Zealanders had seen the virus and the chorus to defeat the virus lost its stuffing.

Evidence from other countries supports the idea of widespread immunity. The very small secondary overseas outbreaks, such as in China and the Australian state of Victoria are further evidence that widespread immunity is growing. If, instead, immunity were sparse, we should expect many further large outbreaks. Other commentators have condemned the low accuracy of Covid-19 tests, however, Roche now has produced a test that has sensitivity and specificity values approaching perfection (100%) that has now got widespread acceptance in Europe. Not even many of our established antibody tests have achieved this.

The philosopher George Santayana reasoned, “those who cannot remember the past are condemned to repeat it.” At this crucial juncture, history indicates that the value of antibody tests and the idea of growing immunity cannot be so easily dismissed. If the virus is more widespread than the genetic tests indicate, we need to urgently reconsider whether or not border closures and social restrictions are really worthwhile.


Video: epidemiologist’s take on Covid-19

Dr. Simon Thornley

  • Deaths due to coronavirus have been exaggerated
  • Mean age of death – 80 years old

Did lockdowns save anyone?

Sweden’s former top state epidemiologist has claimed unless a vaccine is found soon, lockdowns like New Zealand’s won’t prevent any deaths at all – just push them into the future.

Johan Giesecke’s call, published by journal The Lancet, comes the same week a new paper claims lockdowns in hard-hit western Europe haven’t saved a single life at all, which has split opinion among experts.

Sweden has taken a different approach to handling the COVID-19 pandemic than most other countries, deciding against a lockdown of any kind, instead relying on people following social distancing guidelines. As of Monday, it had 26,300 confirmed cases and 3225 deaths – far more than its Scandinavian neighbours, but only a fraction of those seen in Spain, Italy and the UK, which have all implemented lockdowns of various kinds.

10 Reasons to end lockdown

Dr John Lee, Retired Professor of Pathology, writing in The Spectator:

Even if one could understand why lockdown was imposed, it very rapidly became apparent that it had not been thought through. Not in terms of the wider effects on society (which have yet to be counted) and not even in terms of the ways that the virus itself might behave.


…at the start, there was hardly any evidence. Everyone was guessing. Now we have a world of evidence, from around the globe, and the case for starting to reverse lockdown is compelling.


…Covid is not, in fact, an extraordinarily lethal pathogen, just a nasty one, similar to many others.


…our new normal should look very much like our old, perhaps with the addition of some social responsibility in the face of respiratory illness. It is the only way for us to live in the world.


The Lancet: The invisible pandemic

It has become clear that a hard lockdown does not protect old and frail people living in care homes—a population the lockdown was designed to protect.

Neither does it decrease mortality from COVID-19, which is evident when comparing the UK’s experience with that of other European countries.


PCR testing and some straightforward assumptions indicate that, as of April 29, 2020, more than half a million people in Stockholm county, Sweden, which is about 20–25% of the population, have been infected (Hansson D, Swedish Public Health Agency, personal communication). 98–99% of these people are probably unaware or uncertain of having had the infection; they either had symptoms that were severe, but not severe enough for them to go to a hospital and get tested, or no symptoms at all. Serology testing is now supporting these assumptions.

Everyone will be exposed to severe acute respiratory syndrome coronavirus 2, and most people will become infected. COVID-19 is spreading like wildfire in all countries, but we do not see it—it almost always spreads from younger people with no or weak symptoms to other people who will also have mild symptoms. This is the real pandemic, but it goes on beneath the surface, and is probably at its peak now in many European countries. There is very little we can do to prevent this spread: a lockdown might delay severe cases for a while, but once restrictions are eased, cases will reappear. I expect that when we count the number of deaths from COVID-19 in each country in 1 year from now, the figures will be similar, regardless of measures taken.

Should New Zealand be eliminating coronavirus?

24 April.

Simon Thornley

The Government and its health advisers are taking an increasingly hardline against coronavirus, stating that it will be eliminated from our shores. It certainly is desirable, but is it realistic?

New Zealand is one of the only countries in the world to attempt this. Almost alone, we have shifted from agreeing with the international approach of flattening the curve to the objective of either eliminating or eradicating the virus. The latest claim, or clarification, is that the Government’s intention is ‘zero spread’ rather than ‘zero virus’.

We need to consider that the only means of achieving even zero spread are tough social restrictions, only ending when a vaccine is invented and most of the population is vaccinated. Let’s be clear – that means a form of very restricted activity for at least a year.

The Government contends that these are needed because our population is vulnerable to the virus, so the spread must be stopped. It paints a picture that the virus is contained by the current public health measures as well as lockdown, and we are effectively leaping on and isolating each new case.

Evidence emerging in the rest of the world, however, is that this picture of a lockdown-halted virus amongst a defenseless population is inaccurate.

Serological tests from samples of people in New York, Germany and California, in contrast, show that between 4 to 15 per cent of the population have seen the virus, recovered from it, and are now immune. This is a much larger proportion of the population than we have seen from positive swab tests of the virus.

This has important implications. First, it shows that the mortality of the virus is much lower than previously appreciated. Also, it demonstrates why a suppression strategy is better than elimination. China, which is trying to eliminate the virus, is now experiencing a resurgence in cases. The cat is well and truly out of the bag.

To boot, recent analysis from the US shows that lockdowns are not effective in reducing Covid-19 deaths, comparing states with such a policy to those without. The data shows that the strongest factor determining a State’s Covid-19 deaths is population density. The lower it is, the lower the death rate. This is a key factor in New Zealand’s favour.

In New Zealand, until we have some data on existing immunity, we just cannot tell how realistic elimination is. That’s without considering whether the goal is desirable or the means worth the cost.

We are betting the house on something that overseas data is showing to be an increasingly remote possibility. Perhaps the rest of the world knows something we don’t?

Study shows no relationship between lockdowns and lower Covid-19 deaths


Comparing US states shows there is no relationship between lockdowns and lower Covid-19 deaths.